Two research groups from ETH Zurich have developed a method that is able to realistically simulate nanoelectronics devices and their properties quickly and efficiently.
It is believed that this method could be used to provide a solution to overheating problems experienced with the use of increasingly small and powerful transistors used in key applications, for example, data centers where up to 40% of power consumption is used to meet cooling requirements.
With demand for increasingly powerful supercomputers that are even smaller and more powerful, the industry is being driven to assemble transistors that measure in at just a few nanometers. However, increasing levels of heat dissipation are making it difficult for manufacturers to meet these demands whilst keeping energy consumption at efficient levels.
A representation of self-heating in a Fin field-effect transistor (FinFET) at high current densities. Image Credit: Jean Favre, CSCS).
Developing the Simulation Booster
Historically, conventional programming and supercomputers only allowed researchers to simulate heat dissipation in transistors consisting of around only 1,000 atoms. This is because communication between processors and memory made it impossible to simulate larger objects.
In an attempt to make new nano transistors more efficient, the ETH Zurich research group, led by professors Torsten Hoefler and Mathieu Luisier, from the Integrated Systems Laboratory at ETH Zurich simulated transistors using what is known as a ‘quantum transport simulator’, a piece of software named OMEN.
OMEN runs calculations based on what is known as density functional theory. This allows for a realistic simulation of transistors in atomic resolution at the quantum mechanical level. Using this, simulations visualize how electrical currents flow through a nano transistor and how electrons interact with crystal vibrations. This allowed the Zurich researchers to accurately identify sites of heat production.
By being able to see where heat is being produced and dissipated, design engineers are provided with a top-down view that can be used to optimize nano transistors and make them more energy efficient.
Realistic Transistor Simulation
By being able to realistically simulate transistors up to ten times the size of previous limitations, future transistor development stands to benefit hugely.
Using DaCe OMEN—an improved iteration of the original simulation software tool— developers could potentially produce realistic simulations of transistors made up of up to 10,000 atoms and up to 14 times faster than the original method for only 1,000 atoms. This will not only speed up transistor development but also make them far more efficient with optimized design.
The application of DaCe OMEN also revealed that most heat is generated towards the end of the nano transistor channel and that it spreads to affect whole systems.
Earning the two ETH Zurich teams the Gordon Bell Prize, DaCe OMEN is far more efficient than OMEN, able to simulate a realistic transistor 140 times faster with a sustained performance of 85.45 petaflops per second.
The scientists behind DaCe OMEN are convinced that this new process for simulating electronic components has a great many potential applications, one being in the manufacture of Li-ion batteries that can become volatile when they overheat.